CyberBullying Detection in Social Media

• 

  Text Detection

• 

  Image Detection

Inspiration

Social Networking Sites are widely used nowadays. Everyday something or the other is posted as a textual message, images or videos on twitter, Instagram and Facebook. This has also paved the way for online crimes like cyberbullying. Due to increase in cyberbullying, research in this field has achieved milestones. Depending upon different categories content based, sentiment-based, network-based are the areas that are being for research. The various forms of cyberbullying occurring nowadays are sexual, racial and physical-disability based. These include posting of mean-spirited messages, abusive words, sexual videos and personal images of a person.

What it does

It works in real-time. It classifies and detects what user is trying to send to other person. If user is sending message then the app will be check if the message is appropriate or not. Based on model trained if the message is suitable to send , it will allowed else the message will be blocked. Same applies while sending images. If the images are inappropriate then they will be blocked

How I built it

I used python tkinter for GUI and sockets in python to build simple chat app for demonstration purpose. Flair model which uses LSTM is used to detect the text which user is sending. For images Transfer learning in Keras and use MobileNetV2 for classification and detection of images.

Challenges I ran into

Making a real-time application was big challenge for me. Increasing accuracy of classification of images and text is other challenge which i covered up to a greater extent.

Accomplishments that I'm proud of

Making it work in real-time. As soon as user sends messages either text or images it will instantly provide the results.

What I learned

Big thing i learned is how mobilenet can be used to classify images. There are lot of readymade libraries on internet and how we can used them to solve simple simple problems in real life

What's next for CyberBullying Detection in Social Media

Still the project is in development phase , it may fail to detect sometimes. Application takes some time to process messages making the app not respondable for little time, so its another challenge to solve.

Built With

• deep-learning
• flair
• mobilenetv2
• python